Subject: Re: TCP fast retransmit
To: Greg Troxel <gdt@ir.bbn.com>
From: Michael Eriksson <Michael.Eriksson@era-t.ericsson.se>
List: tech-net
Date: 12/15/2004 23:23:41
> If the receive window is updated, one can't infer that the ack was due
> to another arriving packet - it could simply have been generated by
> the TCP receiver reading data.

Ah, of course.

But in this case there was just a very minor update of the window (142
bytes of a 32k window). Maybe the window should be frozen when the
stack is sending dup acks (and the window is already big enough).
Missing the fast retransmit really kills (interactive) performance.

Which leads to the next question: what's up with NetBSD's SACK
support? There's already a net.inet.tcp.sack sysctl variable, and some
traces of initial implementation work, like a SACK option decoder
(which doesn't do anything).

/Micke